IRJET- Secured Hadoop Environment

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 05 Issue: 06 | June-2018 www.irjet.net p-ISSN: 2395-0072
© 2018, IRJET | Impact Factor value: 6.171 | ISO 9001:2008 Certified Journal | Page 478
SECURED HADOOP ENVIRONMENT
Safeeda1, Prof. Savitha C. K2, Prof. Prajna M. R3, Prof. Ujwal U4
1M.tech Student, Dept. of Computer Science, K.V.G College of engineering, Sullia, Karnataka, India
2,3Professor, Dept. of Computer Science, K.V.G College of engineering, Sullia, Karnataka, India
4Head of the Department, Department of Computer Science, K.V.G College of Engineering, Sullia, Karnataka, India
-----------------------------------------------------------------------***--------------------------------------------------------------------
Abstract-Data is growing at an enormous rate in the
present world. One of the finest and most popular
technologies available for handling and processing that
enormous amount of data is the Hadoop ecosystem.
Enterprises are increasingly relying on Hadoop for storing
their valuable data and processing it. However, Hadoop is
still evolving. There is much vulnerability found in Hadoop,
which can question the security of the sensitive
information that enterprises are storing on it. In this
paper, security issues associated with the framework have
been identified. We have also tried to give a brief overview
of the currently available solutions and what are their
limitations. At the end a novel method is introduced, which
can be used to eliminate the found vulnerabilities in the
framework. In the modern era, information security has
become a fundamental necessity for each and every
individual. However, not everyone can afford the
specialized distributions provided by different vendors to
their Hadoop cluster. This paper presents a cost-effective
technique that anyone can use with their Hadoop cluster
to give it high security.
Keywords: Big data, encryption, Hadoop, information
security, Mahout, etc…
1. INTRODUCTION
Big Data is the term that refers not only to the large
volumesof data but it is also concerned about the
complexity of the data and the speed at which it is
getting generated. It is generally described by using
three characteristics,widely known as 3 V's:
 Volume
The size is one of the characteristics that define big data.
Big data consists of very large data sets. However, it
should be noted that it is not the only one parameter and
for data to be considered as big data, other
characteristics must also be evaluated.
 Velocity
The speed at which the data is being generated is an
important factor. For example, in every one second
thousands of tweets are tweeted on micro blogging
platform, Twitter. Even if the size of each individual
tweet is 140 characters, the speed at which it is getting
generated makes it an eligible data set that can be
considered as big data.
 Variety
Big data comprises data in all formats: structured,
unstructured or combination of both. Generally, it
consists of data sets, so complex that traditional data
processing applications are not sufficient to deal with
them. All these characteristics make it difficult for
storing and processing big data using traditional data
processing application software's. Two papers published
by Google [1], [2] build the genesis for Hadoop. Hadoop
is an open source framework used for distributed
storage and parallel processing on big data sets.
Two core components of Hadoop are:
 Hadoop distributed file system (hdfs)
Used for distributed storage of data. The input file is first
split into blocks of equal size except the last block which
are then replicated across Data Nodes. Currently, default
block size is 128 MB which was previously 64 MB and
default replication factor is 3. Block size and replication
factors are configurable parameters.
 Mapreduce
It is used for parallel processing on distributed data on
cluster of commodity hardware in a reliable, fault-
tolerant manner. A MapReduce job usually splits the
input data set intoindependent chunks which are
processed by the map tasks in a completely parallel
manner. The framework sorts the outputs of the map
function, which are then given as input to the reduce
tasks. There are plenty of resources available [3], [5] that
describe the detailed architecture of Hadoop and about
how it works.
Our focus in this paper is on identifying the security
problems that can occur on a typical Hadoop
environment, what are the currently available solutions
for encrypting data in Hadoop and their limitations, and
finally, a novel approach that can be used for large scale
encryption in Hadoop environment in efficient manner
will be discussed.

2. HADOOP SECURITY
To perceive a basic idea about Hadoop security, one has
to look back at the origin of Hadoop. Security was not
exactly the priority for Doug Cutting and his team at the
time they started developing the Hadoop [6]. Initially, it
was just a part of Apache Nutch project until it was
moved to the new Hadoop subproject in January 2006
[7]. A basic authentication level using username and
password is provided in default Hadoop installation.
Many a times, it will not be sufficient. As a default,
Hadoop doesn't give support for any authentication of its
users or Hadoop services. A user only authenticates with
the operating system during the logon process. After this
process, when the user invokes the Hadoop command,
the user id and group is set by executing whoami and
bash -c groups respectively. So if a user writes their own
whoami script and adds it to the path before the Linux
whoami is called, the user should beable to impersonate
any user including the super user in the Hadoop file
system.
 Security of data in motion
In Hadoop, actual data is stored in distributed manner on
multiple DataNodes while NameNode stores the
metadata and edit log. The actual communication of data
blocks happens between Client Node and Data Node.
Hence, Hadoop framework consists of several nodes with
data communication between them. The data is
transmitted over the network, which is not encrypted by
default. The data which is being transmitted is therefore
open to attack by hackers. Various communication
protocols such as Remote Procedure Call (RPC),
Transmission Control Protocol over Internet Protocol
(TCP/IP), and Hypertext Transfer Protocol (HTTP) are
used for internode communication. There are some
solutions available for securing communication between
various nodes like Kerberos and Simple Authentication
and Security Layer (SASL) which are discussed in next
section.
 Security of data at rest
Data at rest refers to data stored in persistent storage. By
default Hadoop doesn't encrypt data that's stored on
disk and that can expose sensitive data to security
attacks. This is especially a big issue due to the nature of
Hadoop architecture, which spreads data across a large
number of nodes; exposing the data blocks at all those
unsecured entry points. There are a number of choices
for implementing encryption at rest with Hadoop. One of
them is to use encryption zone a new layer of abstraction
to HDFS. It will be discussed along with its limitations in
next section.
We have to also understand that since Hadoop usually
deals with large volumes of data and
encryption/decryption takes time, it is important that
the framework used performs the
encryption/decryption fast enough, so that it doesn't
impact performance.
Today, Hadoop is no more a tool for experimentation.
Rather, it is getting increasing popularity at enterprise
level. It definitely gives those reliable and cost-effective
big data storage and processing platform and an obvious
competitive advantage. But along with that there are
some risks associated with it. For example, there is a risk
of data leakage in transit while it is getting transferred
over network from Hadoop client to DataNode.
There are some Hadoop distributors like IBM, Cloudera
and Horton works that claims to be providing security to
the client's data. Even if their claims are true, not
everyone can afford to use a specialized distribution.
Getting information security has now become
fundamental right in modern era. There should be some
open frameworks which anyone canuse easily and get a
highly secure Hadoop environment. Also, even if
someone is using the services provided by Hadoop
distributors, they may have to face some problems. The
sensitive data of enterprises is stored on cloud and all
the services are accessed through Internet. Since, the
data is stored on cloud; companies have to face some
security and privacy challenges:
 Multi-tenancy
Cloud providers have to build an infrastructure that is
cost effective and efficiently scalable to meet customer's
requirement. In order to do so they need share storage
devices andphysical resources between multiple users.
This is called multi-tenancy. But sharing of resources
means that attackers can easily target another customer
if they both are using the same physical devices and
proper security measures are not implemented.
 Loss of control
As the companies do not have direct control over their
data, they can never know if their data is being used by
someone else. This can lead to security issues since there
are no transparent mechanisms to monitor the resources
directly. There is also a possibility that their data is not
completely removed at the time they discontinue using
services provided by those service providers.
 Trust chain in clouds
As described earlier, customers have to share physical
resources with other customers and they do not have
direct control over their data. Therefore, customers rely
on the cloud providers using trust mechanisms as an
alternative to giving users transparent control over their
data and cloud resources. Cloud providers needs to build

confidence amongst their customers by assuring them
that the provider's operations are certified in compliance
with organizational safeguards and standards.
 Privacy concerns
Privacy and Security are two distinct domains. Yet they
are usually discussed together since security is required
in order to provide privacy. The enterprises needs to be
sure that their sensitive data is not being accessed by
cloud providers and that it is not being share with some
third party in return for some money.
There is one more reason for why an organization
should secure its Hadoop environment. Security and
privacy standards such as International Organization for
Standardization (ISO) have evolved. This requires
service providers to comply with these regulatory
standards to fully safeguard their client's data assets.
This has resulted in very protective data security
enforcement within enterprises including service
providers as well as the clients.
We have discussed some of the major issues that are
associated with the use of Hadoop as the big data storage
and processing system in this section and why an
enterprise needs to secure its Hadoop environment. One
of the solutions to overcome those challenges is to
encrypt the data. So even if encrypted data in motion is
captured by some unauthorized person or it is stored at
some remote servers, it will still be safe since
unauthorized party will not be able to interpret the
actual meaning of the data. Some of the present solutions
to achieve security in Hadoop environment and their
limitations are discussed in next section.
3. AVAILABLE SOLUTIONS
There are some security measures available for securing
sensitive data sets in Hadoop, while they are residing in
Hadoop or while they are transferred across the
network. In any distributed system, when two parties
(the client and server) have to communicate over the
network, the first step in this communication is to
establish trust between these parties. This is usually
done through the authentication process, where the
client presents its password to the server and the server
varies this password. If the client sends passwords over
an unsecured network, there is a risk of passwords
getting compromised as they travel through the network.
Kerberos is a secured network authentication protocol
that provides strong authentication for client/server
applications without transferring the password over the
network. Kerberos works by using time-sensitive tickets
that are generated using the symmetric key
cryptography. Kerberos is derived from the Greek
mythology where Kerberos was the three-headed dog
that guarded the gates of Hades.
The three heads of Kerberos in the security paradigm
are:
1) The user who is trying to authenticate.
2) The service to which the client is trying to
authenticate.
3) Kerberos security server known as Key Distribution
Center (KDC), which is trusted by both the user and
the service. The KDC stores the secret keys
(passwords) for the users and services that would
like to communicate with each other.
 Encryption for data in motion
To protect the data in motion, it is required to
understand the underlying protocol that is used when
data is transferred over the network in Hadoop. A
Hadoop client connects to NameNode using the Hadoop
RPC protocol over TCP, while the Hadoop client transfers
the data to DataNode using the HTTP protocol over TCP.
User authentication to Name Node and JobTracker
services is through Hadoop's remote procedure call
using the SASL framework. Kerberos as discussed
previously is used as the authentication protocol to
authenticate the users within SASL. The SASL
authentication framework can be utilized to encrypt the
data while it is being transported into the Hadoop
environment thereby protecting the data in motion. SASL
security guarantees that data exchanged between the
client and servers is encrypted and is not readable by a
``man in the middle''. SASL encryption can be enabled by
configuring the property hadoop rpc protection to
privacy in core-site.xml. This ensures that the
communication between the Hadoop client and
NameNode is secured and encrypted.
Any Hadoop client requesting for data from HDFS needs
to fetch the data blocks directly from DataNode after it
fetches the block ID from NameNode. The Block Access
Token (BAT) can be used to ensure that only authorized
users are able to access the data blocks stored in
DataNodes. There is an excellent compilation of various
security mechanisms available for securing Hadoop in
[8].
 Encryption for data at rest
There are a number of choices for implementing
encryption at rest with Hadoop. One of them is using the
encryption zone. For transparent encryption, Hadoop
has introduced a new abstraction to HDFS: the
encryption zone [9]. An encryption zone is a special
directory whose contents will be encrypted
transparently upon write and transparently decrypted
upon read. Here, every encryption zone is composed of
single encryption zone key which is specified when the

zone is formed. Each file residing within an encryption
zone has its own unique data encryption key (DEK).
DEKs are never handled directly by HDFS. HDFS only
handles an encrypted data encryption key (EDEK). The
Client will perform decrypting an EDEK, and then it will
make use of the subsequent DEK to read and write data.
HDFS DataNodes simply see a stream of encrypted bytes.
A new cluster service should be there to manage
encryption keys which is known as the Hadoop Key
Management Server (KMS).
In the process of HDFS encryption, the KMS performs
three basic responsibilities:
1) Providing the access permission to stored
encryption zone keys.
2) Generating new encrypted data encryption keys for
storage on the Name Node.
3) Decrypting encrypted data encryption keys for use
by HDFS clients. As one can easily identify, this
process is similar to what we have seen about
Kerberos actual key is never transmitted over
network for improved security. For more details
about encryption zone and how it can be used to
encrypt data at rest, refer [10], [11].
In this section we have discussed about how Kerberos
can be used to authenticate users securely and how SASL
can be used to encrypt data in transit. We have also
learned about encryption zone which is implemented
currently in Hadoop for storing encrypted data at rest. So
far it seems that Hadoop already have what it takes to
give its users a completely secure environment.
However, there are some limitations associated with
them: Most of times it is not sufficient to just store the
data but we must also be able to process the data in
efficient manner. The Hadoop encryption zone is not
suitable for processing. If we want to run Map Reduce
task on data stored in encryption zone then we first have
to decrypt the complete file and make it available for
Map Reduce task in the decrypted form. Because Hadoop
usually deals with large volumes of data and
encryption/decryption takes time, it is important that
the framework used performs the
encryption/decryption fast enough that itdoesn't impact
performance. Hadoop encryption zone uses cipher suit
like Advanced Encryption Standard (AES) which is good
encryption standard but it is definitely having higher
memory requirement and can degrade performance
since Client node is having limited memory and files
used are generally of larger size.
4. PROPOSED FRAMEWORK
In this section we will propose a framework which
canhelp eliminate limitations discussed in previous
section. This framework does not intend to replace the
existing solutions but rather it tries to add flexibility in
the current process. This can be added as an additional
process that can be usedalong with Kerberos and SASL to
give all three-dimensional security to any Hadoop
cluster. Three-dimensional security is nothing but the
term coined by authors in order to show that by using
the proposed framework, anyone can get assured about
security of their data in following three ways at the same
time:
1) Secure user authentication.
2) Encrypted data in transit.
3) Encrypted data at rest.
One of the major benefits of using this framework is that
Map Reduce tasks can directly be executed on data
stored in HDFS without requiring the complete file to be
decrypted before it can be processed and hence it
completely eliminates the use of HDFS encryption zone.
However, it should be noted that individual data blocks
stored in MapReduce are decrypted at the time of
Mapping but that process is faster and memory efficient
since size of individual data blocks is slow as compared
to the original file. It should also be noted that HDFS
encryption zone can still be used if data to be stored
there will not frequently be required to be accessed for
longer period of time.
The Hadoop client who initiates data read-write
operation is also responsible for splitting the complete
data set into blocks of size as specified in Hadoop
configuration file. Currently the default block size is 128
MB which was previously 64 MB. In the proposed
system, the Hadoop Client encrypts each individual block
before it is stored into HDFS. This ensures that
MapReduce can process each block independently, and
the decryption logic is applied during the map phase for
a MapReduce job. The decryption key should be made
available to the MapReduce job to decrypt the file.
This is provided to the MapReduce program through the
job configuration. The solution isscalable and efficient.
The solution is most effective if used along with
Kerberos and SASL for secure RPC communication and
encrypting data in motion. This combination provides
the Three-Dimensional Security in Hadoop environment.
This approach is shown in the following fig. 1.

Fig -1: Providing Three-Dimensional Security in Hadoop
Environment.
To support block-level encryption in Hadoop, the client
should encrypt each individual block before it is
transferred to DataNode for storing. MapReduce jobs can
directly be executed on encrypted data on HDFS
provided that decryption key is known to Mapper class.
Similarly, when the file has to be read or copied to local
machine, the client will read each individual blocks and
then apply the decryption logic in the client side.
An important observation to be made here is that the
encryption credentials are not stored in the Hadoop
cluster. The encryption credentials are completely
managed securely on the client side using some KMS like
Kerberos. Now the question arises that what is the
algorithm that will be used to encrypt each individual
block. This is the most critical part of the process. The
algorithms chosen must be memory efficient, faster and
secure. Also, Kuber framework is not dependent on a
single encryption technique but rather gives client the
flexibility to implement their own encryption algorithms
based on their needs. Two of the most reliable
candidates are AES and Salsa20. The current
implementation of the Kuber uses a variant of Salsa20
calledChaCha20 because of the many reasons. Note that
ChaCha20 was designed in order improve Salsa20
specifically in terms of diffusion per round and
increasing resistance to cryptanalysis, while preserving
time per round. Some of the reasonsto choose ChaCha20
instead of AES with initial version of Kuber are:
An efficient implementation of Salsa20 can be as much as
3-5 times faster than AES. ChaCha20 being a variant of
Salsa20 gives the same or sometime better speed
benefits as Salsa20.
One of the most common attacks in symmetric key:
 Cryptography is differential cryptanalysis. Salsa20
and its variants like ChaCha20 have been proved to
be secure against differential cryptanalysis.
 ChaCha20 doesn't require any lookup tables and
avoids the possibility of timing attacks
 AES is somewhat complex and troublesome to
implement in software. On the other hand, ChaCha20
is easy implement.
 In order to make AES run fast one has to expand the
key and pre-compute a series of tables all of these
processes increases key setup time and potentially
makes vulnerable to cache timing attacks.
 An advantage with block ciphers like AES is that they
can randomly access just a portion of encrypted
message without decrypting the complete message.
Internally, chacaha20 works like a block cipher used
in counter mode.
 Consequently it can also be used to randomly access
any block of generated key stream. Google's security
team has already added the suites to Open SSL.
 The Mozilla team is working on adding it to Firefox.
Google also encouraging the increased adaption of
the cipher suite in order to offer safer & faster
alternatives.
 Salsa20 and its variants like ChaCha20 are free for
any use.
5. ALGORITHM DESCRIPTION
ChaCha20 stream cipher is a variant of the Salsa20
family that follows the same design principles as Salsa20
with some changes in the details, most importantly to
increase theamount of diffusion per round. The authors
also claim that the minimum number of secure rounds
for ChaCha20 is smaller than the min imum number of
secure rounds forSalsa20. It takes a key (k) of length 32
bytes (256 bits) and a nonce (r) of length 8 bytes (64
bits) as input along with the plaintext. Nonce is a unique
value that is never going to change as long as the key is
fixed. The pair (k, r) is never used more than once. So we
can reuse the key since pair (k, r) is unique. The 256-bit
key is then expanded into 264 randomlyaccessible
streams, each containing 264 randomly accessible 64-
byte blocks. It encrypts a b-byte plaintext by XOR'ing the
plaintext with the first b bytes of the stream and
discarding the rest of the stream. Decryption operation is
same as encryption as a b-byte cipher text is XOR'ed with
first b bytes of the stream. This is a common scenario
that you will observe with stream ciphers. The quarter
round of ChaCha20 is executed as follows.
The algorithm updates four 32-bit state words a, b, c, d
as follows:
a C D b; d ^ D a; d <<<D 16;
cC D d; b ^ D c; b <<<D 12;
a C D b; d ^ D a; d <<<D 8;
cC D d; b ^ D c; b <<<D 7;
This code is expressed in common extension of C
language. ^means XOR, + means addition modulo 232,

and <<<b means b-bit rotation of a 32-bit integer
towards high bits. We can observe here that each word is
updated twice in a quarter round of ChaCha20.
As we see, ChaCha20 is very easy to implement, faster on
all types of CPUs giving the same level of security as the
other standard encryption algorithms such as AES. Also,
it should be noted that Kuber framework is not limited to
any specific encryption algorithm. Developers have the
flexibility to use their own implementation of encryption
systems they are comfortable with.
 Analysis using mahout environment:
The Apache Mahout offers a library which includes
scalable machine-learning algorithms, which are
implemented on top of Apache Hadoop by using the
MapReduce paradigm. Machine learning is a branch of
artificial intelligence which concentrates on permitting
machines to learn without being explicitly programmed,
and it is commonly used to improve future performance
based on previous outcomes.
After big data is kept on the Hadoop Distributed File
System (HDFS), Mahout will provide the data science
tools in order to automatically find meaningful patterns
in those big data sets. The aim of the Apache Mahout
project is to make this process faster and easier which
will turn big data into big information.
 What Mahout Does
Mahout provisions four main data science use cases
which are listed below:
Collaborative filtering – It mines user behavior and
outputs product recommendations (e.g. Amazon
recommendations).
Clustering – It will collect items in a particular class
(such as web pages or newspaper articles) and then it
will organize them into naturally occurring groups, such
that items which belongs to the same group are similar
to each other.
Classification – It will learn from existing
categorizations and then allocates unclassified items to
the best category.
Frequent itemset mining – It analyzes items in a group
(e.g. items placed at shopping cart or terms in a query
session) and then identifies which items will appear
together.
 Features of Mahout
 The algorithms of Mahout are written on top of
Hadoop. Hence, it works fine in distributed
environment. Mahout utilizes the Apache Hadoop
library in order to support scalability effectively in
the cloud.
 Mahout provides the coder a ready-to-use
framework for performing data mining tasks over
large volumes of data.
 Mahout allows applications to analyze large sets of
data effectively and in less time.
 It consists of numerous MapReduce facilitated
clustering implementations for example k-means,
fuzzy k-means, Canopy, Dirichlet, and Mean-Shift.
 Supports Distributed Naive Bayes and
Complementary Naive Bayes classification
implementations.
 It provides distributed fitness function
capabilities for evolutionary programming.
 Includes matrix and vector libraries.
 Applications of Mahout
 The Companies which includes Adobe, LinkedIn,
Facebook, Twitter, Foursquare and Yahoo are
using Mahout internally.
 Foursquare-this helps in finding out places, food,
and entertainment that are offered in a
particular area. It utilizes the recommender
engine of Mahout.
 Twitter makes use of Mahout for user interest
modelling.
 Yahoo! utilizes Mahout for pattern mining.
 Applying Encryption over data to be clustered:
In this work the data to be clustered is encrypted before
uploading to HDFS. Then the data is uploaded to HDFS in
encrypted format.The encrypted data is decrypted
before clustering. The clustering is performed after this
process and the time taken for the encryption process is
analyzed. Thus it ensures that the data to be clustered
will be securely stored in Hadoop environment.
6. RESULTS AND DISCUSSION
Having decided to use ChaCha20, We started to
experiment with the speed of the algorithm with files of
different sizes and different buffersizes which will be
given as an input to the algorithm. The results are noted
in Table 1.

Table -1: Speed (Mb/s) of encryption with various file
sizes' (rows) and buffer sizes' (columns).
Configuration: Intel Core i3 CPU with 4 GB RAM running
on Windows 8.1.
The experiment started with an assumption that
increasing the buffer size will reduce the iterations of
loop and consequently reduce the overall encryption
time. However, the results were somewhat different. The
highest speed was achieved with buffer size between
500 KB and 1 MB. It is because the cost of machine cycles
associated with transferring larger blocks like 64 MB
from one method to other isgreater than the time saved
by reducing the number of iterations of the loop. Fig.2.
shows the graphical representation of the results. The
highest speed in each case is achieved when buffer size
of 500 KB or 1 MB is used.
Chart -1: Encryption speed in Mb/s of 6 files with
various buffer sizes.
ChaCha20 is designed such that it can encrypt any
random block of input data, making it suitable to be used
as a block cipher. Another major advantage of this
method is that each encrypted block can be individually
decrypted provided you have the key and nonce for the
decryption. Consequently, MapReduce tasks can be
executed directly on encrypted data stored in HDFS. The
complete process for large scale encryption in Hadoop
environment can be summarized as follows:
The Hadoop client configures the key and nonce for
encryption process in KMS. Data to be stored in HDFS is
first encrypted at client side by dividing it into smaller
chunks using parallel computing.
MapReduce tasks can be directly executed on encrypted
data. The level of flexibility can be imagined by the fact
that any current implementation of MapReduce can be
adapted to use Kuber with only two lines of modification
in their original program. All they have to do is to import
corresponding decryption library and call decryption
method before processing the data. Authorized clients
can access the encrypted file stored in HDFS. Decryption
process is similar to encryption process.
7. CONCLUSION
Data is growing at very high speed in the present world.
One of the finest and most popular technologies available
for handling and processing that enormous amount of
data is the Hadoop ecosystem. Enterprises are
increasingly relying on Hadoop for storing their valuable
data and processing it. However, big data exposes the
enterprise to numerous data security threats. In this
work encryption and decryption technology is applied
over data before uploading to the HDFS, thereby
providing security. The Chacha20 algorithm is used for
this process. The analysis is done using mahout
environment over the data to be clustered. The results
are observed. Thus this work presents an effective
technique for securing the big data to be clustered in
hadoop Environment.
REFERENCES
1. S. Ghemawat, H. Gobioff, and S.-T. Leung, ``The
Google _le system,'' in Proc. 19th ACM Symp. Oper.
Syst. Principles (SOSP), 2003, pp. 29_43.
2. S. Ghemawat and J. Dean, ``MapReduce: Simpli_ed data
processing on large clusters,'' ACM Commun. Mag., vol.
51, no. 1, pp. 107_113, Jan. 2008.
3. D. Borthakur, ``The Hadoop distributed file system:
Architecture anddesign,'' Hadoop Project Website, vol.
11, p. 21, Aug. 2007.
4. T. White, Hadoop: The De_nitive Guide. Farnham,
U.K.:O'Reilly, 2012.
5. D. de Roos, P. C. Zikopoulos, R. B. Melnyk, B. Brown,
and R. Coss, Hadoop For Dummies. Hoboken, NJ, USA:
Wiley, 2014.

6. B. Lakhe, Practical Hadoop Security. New York, NY,
USA: Apress, 2014, pp. 1946.
7. Apache Hadoop, accessed Dec. 2016. [Online].
Available: https://en.wikipedia.org/wiki/Apache
Hadoop History.
8. P. P. Sharma and C. P. Navdeti, ``Securing big data
Hadoop: A review of security issues, threats and
solution,'' Int. J. Comput. Sci. Inf. Technol.,vol. 5, no. 2,
pp. 2126_2131, 2014.
9. Apache Hadoop 2.7.3_Transparent Encryption in
HDFS, accessed Feb. 2017. [Online]. Available:
https://hadoop.apache.org/docs/r2.7.2/ hadoop-
project-dist/hadoop-dfs/TransparentEncryption.html.
10. HDFS Data At Rest Encryption, accessed Feb. 2017.
[Online].Available:https://www.cloudera.com/docum
entation/enterprise/5x/topics/cdh_sg_hdfs_encryptio
n.html.
11. HDFS Encryption Overview, accessed Feb. 2017.
[Online].Available:https://docs.hortonworks.com/HD
PDocuments/HDP2/HDP2.3.2/bk_hdfs_admin_tools/c
ontent/hdfs-encryption-overview.html.

IRJET- Secured Hadoop Environment

Related slideshows

More Related Content

IRJET- Secured Hadoop Environment